Farah Deeba

Unsupervised abnormality detection using saliency and Retinex based color enhancement

An efficient and automated abnormality detection method can significantly reduce the burden of screening of the enormous visual information resulting from capsule endoscopic procedure. As a pre-processing stage, color enhancement could be useful to improve the image quality and the detection performance. Therefore, in this paper, we have proposed a two-stage automated abnormality detection algorithm. In the first stage, an adaptive color enhancement method based on Retinex theory is applied on the endoscopic images. In the second stage, an efficient salient region detection algorithm is applied to detect the clinically significant regions. The proposed algorithm is applied on a dataset containing images with diverse pathologies. The algorithm can successfully detect a significant percentage of the abnormal regions. From our experiment, it was evident that color enhancement method improves the performance of abnormality detection. The proposed algorithm can achieve a sensitivity of 97.33% and specificity of 79%, higher than state-of-the-art performance.

Efficacy Evaluation of SAVE for the Diagnosis of Superficial Neoplastic Lesion

The detection of non-polypoid superficial neoplastic lesions using current standard of white light endoscopy surveillance and random biopsy is associated with high miss rate. The subtle changes in mucosa caused by the flat and depressed neoplasms often go undetected and do not qualify for further investigation, e.g., biopsy and resection, thus increasing the risk of cancer advancement. This paper presents a screening tool named the saliency-aided visual enhancement (SAVE) method, with an objective of highlighting abnormalities in endoscopic images to detect early lesions. SAVE is a hybrid system combining image enhancement and saliency detection. The method provides both qualitative enhancement and quantitative suspicion index for endoscopic image regions. A study to evaluate the efficacy of SAVE to localize superficial neoplastic lesion was performed. Experimental results for average overlap index >0.7 indicated that SAVE was successful to localize the lesion areas. The area under the receiver-operating characteristic curve obtained for SAVE was 94.91%. A very high sensitivity (100%) was achieved with a moderate specificity (65.45%). Visual inspection showed a comparable performance of SAVE with chromoendoscopy to highlight mucosal irregularities. This paper suggests that SAVE could be a potential screening tool that can substitute the application of burdensome chromoendoscopy technique. SAVE method, as a simple, easy-to-use, highly sensitive, and consistent red flag technology, will be useful for early detection of neoplasm in clinical applications.

Automated Growcut for segmentation of endoscopic images

Capsule endoscopy (CE), introduced as a modality for non-invasive examination of entire gastrointestinal tract, demands for an efficient computer-aided decision making system to relieve the physician from the responsibility of screening around 60,000 video frames per patient. An automatic and robust segmentation algorithm can aid the automation of CE screening and decision making procedure. In this paper, we propose a new segmentation algorithm based on GrowCut and apply the algorithm for CE images containing bleeding. To substitute the manual seed input in traditional GrowCut segmentation, the proposed Automated GrowCut (AGC) algorithm initially segments the images using clustering. The cluster centroids, subsequently labeled as bleeding, non-bleeding and background by a trained SVM classifier, serve as seeds for the GrowCut segmentation. A comprehensive evaluation and comparison with respect to ground truth exhibits that the proposed method can achieve a Dice Similarity Coefficient of 0.81, comparable to the interactive GrowCut, requiring only 13.96% of the computation time of interactive GrowCut. The comparison with two other state-of-the-art unsupervised segmentation methods, unsupervised GrowCut and Fuzzy c-means, further justifies the suitability of the proposed method for automated segmentation and annotation of CE images.

Application of modified ant colony optimization for computer aided bleeding detection system

Wireless capsule endoscopy (WCE) plays a significant role in the non-invasive small intestine screening for obscure gastrointestinal bleeding detection. However, the task of reviewing 60,000 frames to detect the bleeding encumbers the clinician, leading to visual fatigue and false diagnosis. In this paper, we propose a color feature based bleeding detection system with feature selection using a modified ant colony optimization (MACO) algorithm. We have utilized the feature selection capability of MACO algorithm to find the optimum feature subset over the color space of RGB and HSV, which provided a classifier that outperforms the classifier formed from RGB and HSV features individually. Comprehensive experimental results reveal that the proposed MACO algorithm can detect the optimal feature subset with performance comparable to exhaustive search in case of individual classifier from RGB and HSV requiring 2% of the computational time compared to exhaustive search. The comparative study of feature selection showed that MACO can provide the most relevant features and improve the performance in terms of accuracy, sensitivity and computational time.

Tri-Scan: A Three Stage Color Enhancement Tool for Endoscopic Images

Modern endoscopes play a significant role in diagnosing various gastrointestinal (GI) tract related diseases where the visual quality of endoscopic images helps improving the diagnosis. This article presents an image enhancement method for color endoscopic images that consists of three stages, and hence termed as “Tri-scan” enhancement: (1) tissue and surface enhancement: a modified linear unsharp masking is used to sharpen the surface and edges of tissue and vascular characteristics; (2) mucosa layer enhancement: an adaptive sigmoid function is employed on the R plane of the image to highlight micro-vessels of the superficial layers of the mucosa and submucosa; and (3) color tone enhancement: the pixels are uniformly distributed to create an enhanced color effect to highlight the subtle micro-vessels, mucosa and tissue characteristics. The proposed method is used on a large data set of low contrast color white light images (WLI). The results are compared with three existing enhancement techniques: Narrow Band Imaging (NBI), Fuji Intelligent Color Enhancement (FICE) and i-scan Technology. The focus value and color enhancement factor show that the enhancement level achieved in the processed images is higher compared to NBI, FICE and i-scan images.

An empirical study on the effect of imbalanced data on bleeding detection in endoscopic video

In biomedical applications including classification of endoscopic videos, class imbalance is a common problem arising from the significant difference between the prior probabilities of different classes. In this paper, we investigate the performance of different classifiers for varying training data distribution in case of bleeding detection problem through three experiments. In the first experiment, we analyze the classifier performance for different class distribution with a fixed sized training dataset. The experiment provides the indication of the required class distribution for optimum classification performance. In the second and third experiments, we investigate the effect of both training data size and class distribution on the classification performance. From our experiments, we found that a larger dataset with moderate class imbalance yields better classification performance compared to a small dataset with balanced distribution. Ensemble classifiers are more robust to the variation in training dataset compared to single classifier.In biomedical applications including classification of endoscopic videos, class imbalance is a common problem arising from the significant difference between the prior probabilities of different classes. In this paper, we investigate the performance of different classifiers for varying training data distribution in case of bleeding detection problem through three experiments. In the first experiment, we analyze the classifier performance for different class distribution with a fixed sized training dataset. The experiment provides the indication of the required class distribution for optimum classification performance. In the second and third experiments, we investigate the effect of both training data size and class distribution on the classification performance. From our experiments, we found that a larger dataset with moderate class imbalance yields better classification performance compared to a small dataset with balanced distribution. Ensemble classifiers are more robust to the variation in training dataset compared to single classifier.

Performance assessment of a bleeding detection algorithm for endoscopic video based on classifier fusion method and exhaustive feature selection

Abstract Capsule Endoscopy (CE) is a non-invasive clinical procedure that allows examination of the entire gastrointestinal tract including parts of small intestine beyond the scope of conventional endoscope. It requires computer-aided approach for the assessment of video frames to reduce diagnosis time. This paper presents a computer-assisted method based on a classifier fusion algorithm which combines two optimized Support Vector Machine (SVM) classifiers to automatically detect bleeding regions present in CE frames. The classifiers are based on RGB and HSV color spaces; the image regions are characterized on the basis of statistical features derived from the first-order histogram probability of respective color channels. A nested cross validation strategy has been adopted for the parameter tuning and feature selection to optimize the classifiers. The optimum feature sets for the best performance are evaluated after exhaustive analysis. The proposed fusion approach achieves an average accuracy of 95%, sensitivity of 94% and specificity of 95.3% for a dataset of 8872 CE frames, which is higher than that obtained from a single classifier. Comparison with the state-of-the-art algorithms exhibits that the proposed method yields superior performance for diverse dataset.

Are Current Advances of Compression Algorithms for Capsule Endoscopy Enough? A Technical Review

The recent technological advances in capsule endoscopy system have revolutionized the healthcare system by introducing new techniques and functionalities to diagnose gastrointestinal tract. These techniques improve diagnostic accuracy and reduce the risk of hospitalization. Although many benefits of capsule endoscopy are known, there are still limitations including lower battery life, higher bandwidth, poor image quality and lower frame rate, which have restricted its wide use. In order to solve these limitations, the importance of a low-cost compression algorithm, that produces higher frame rate with better image quality and yet consumes lower bandwidth and transmission power, is paramount. While several review papers have been published describing the capability of capsule endoscope in terms of its functionality and emerging features, an extensive review on the compression algorithms from past and for future applications is still of great interest. Hence, in this review, we aim to address the issue by exploring the characteristics of endoscopic images, analyzing the strengths and weaknesses of useful compression techniques, and making suggestions for possible future adaptation.

Feature selection using modified ant colony optimization for wireless capsule endoscopy

In this study, a modified ant colony optimization algorithm has been proposed to find a feature subset most relevant to the classification task. The algorithm incorporates a new heuristic information component based on classification accuracy. The proposed methodology has been applied in a multiclass classification problem of capsule endoscopic images, where image regions will be classified as bleeding, non-bleeding and uninformative regions. 75 dimensional features extracted from five color spaces have been investigated in the experiments. The proposed MACO algorithm efficiently finds the optimum feature subset over the five color spaces including RGB, HSV, Lab, YCbCr, and CMYK, resulting in a feature subset outperforming those obtained individually from each color space. The comparative study with state-of-the-art methods of feature selection demonstrated that MACO can provide the most relevant features and improve the performance in terms of accuracy, sensitivity and computational time.

Learning from imbalanced data: A comprehensive comparison of classifier performance for bleeding detection in endoscopic video

Imbalanced data is an inevitable problem in many real world problems, including bleeding detection from endoscopic videos with a fewer clinically significant examples outnumbered by normal examples. In this paper, we have presented a comprehensive analysis of six different classifier performance for different class distribution of training dataset. We have addressed two questions: 1. Is there any advantage of using a certain classifier over others? 2. For bleeding detection problem, what is the optimal range of class distribution in training data set? We have built seven different training sets with different class distributions to answer the above questions. Besides the standard performance metrics, we have defined a metric to measure the robustness of the classifiers to get the optimal range of class distribution for a certain classifier. From our experiments, we found that balanced training set yields the best performance for all classifiers. Ensemble classifiers are more robust to the variation in training dataset compared to single classifier.